The Impact of Remittances on Exchange Rate Volatility and Inflation Volatility in Dollarized Economies

Transcription

1 The Impact of Remittances on Exchange Rate Volatility and Inflation Volatility in Dollarized Economies Working Paper Helena Glebocki Keefe October 13, 2014 Abstract Remittances provide a stable and growing source of foreign currency for many small developing economies. This research finds that the impact of remittance inflows on the size of short term movements in exchange rates depends on the level of dollarization and development in the economy. Countries with high levels of partial dollarization are vulnerable to drastic exchange rate fluctuations that can have destabilizing effects on the local financial system and domestic economy. Their vulnerability is derived from weak institutions and credibility of policymakers, currency mismatch risk on balance sheets of banks, and a high exchange rate pass through. In addition, by supplying a steady source of foreign currency, remittances provide a cushion that mitigates volatility derived from sudden stops or reversals in other capital flows. With monthly and quarterly data for fifteen developing economies, I conduct a panel analysis to test the linear and non-linear relationship between volatility measures and remittances, using an interaction term between remittances, other capital flows, and development conditions. 1 Introduction Remittances represent a stable inflow of foreign currency for many small developing economies. For some, remittances are now a greater source of foreign currency than foreign direct investment and portfolio investment flows. In this segment of research, I determine whether the impact of remittances on exchange rate volatility depends on the level of dollarization and development in the economy. Additionally, I examine whether by representing a steady inflow of foreign currency, remittances are able to mitigate volatility in times when other forms of capital flows are diminishing. Small open economies are vulnerable to fluctuations in capital flows that may lead to volatility in real and nominal exchange rates. Combes, Kinda and Plane (2011) find that significant increases in such flows may cause the financial system to be more fragile and lead to an appreciation of the real exchange rate. Volatility in exchange rates has a negative impact on growth by decreasing trade volume, investment spending, and profitability (Jorge and Habermeier, 2004). Understanding how remittances impact the volatility in exchange rates is an important step forward in establishing the role remittances play in financial development. In economies with high levels of partial dollarization, dependence on both foreign and local currency may translate to greater sensitivity of exchange Please do not cite without author s permission Department of Economics at Fordham University New York. 441 East Fordham Road, Dealy Hall, Office E543, Bronx, NY 10458, USA. Phone: +1 (718) , fax: +1 (718) , hglebocki@fordham.edu. 1

2 rates to changes in foreign currency inflows. There are three factors that explain this vulnerability and how the effects of remittance inflows may differ between countries with high and low levels of dollarization. First, exchange rate and inflation movements are closely linked in highly dollarized economies, making them more responsive to changes in foreign currency inflows. Alvarez-Plata and Garca-Herrero (2007) attribute the stronger exchange rate pass-through in dollarized economies to more non-traded goods being priced in foreign currency. This subjects a broader set of goods to exchange rate variations. Remittance inflows denominated in foreign currency may exert a stronger influence on exchange rate and price movements in highly dollarized economies because the household can choose to consume or save in either foreign or local currency. Next, dollarization stems from periods of high inflation and a loss in credibility of policy makers (Agenor, 2004). With weak institutions, these economies may be more sensitive to external shocks that lead to volatility. Kose, Prasad and Taylor (2009) argue that strong institutional quality is critical in explaining why some countries benefit from capital inflows, while others face a heightened risk of economic crises. Highly dollarized economies may be vulnerable to variations in foreign currency inflows that cause greater exchange rate and inflation volatility due to the weaker institutions and lack of credibility inherent in the economy. Third, currency mismatch risk on balance sheets of financial institutions and businesses is heightened in partially dollarized economies. Such risk increases the sensitivity of the small open economy to changes in foreign currency inflows which, in turn, influence exchange rates. Volatile inflows can lead to balance sheet risks and bank failures (IMF, 2010). Therefore, dollarized economies are more sensitive to factors that affect fluctuations in exchange rates and prices, which may exacerbate currency mismatch risk. In this research, I find that remittances, like all capital flows, have the potential to increase exchange rate volatility in economies with high levels of partial dollarization. However, the level of institutional quality, financial development, and trade openness alter the properties of remittances and how the inflows affect volatility. Meeting minimum levels of development improves the ability of an economy to absorb and adjust to changes in foreign currency inflows. As an economy gets closer to threshold levels of development, its ability to absorb changes in foreign currency inflows improves. Exchange rate and price variations become less sensitive to changes in inflows as a result. Kose et al. (2009) find that meeting minimum levels of development will decrease the risk of crisis related to volatile inflows and promote economic growth when opening to global financial markets. How the level of development alters the effect of remittances on exchange rate volatility is an important consideration to fully grasp the role remittances play in emerging markets and developing economies. Additionally, in times when other capital flows are diminishing, due to sudden stops or reversals, I find that remittances mitigate the exchange rate volatility derived from the outflow. Remittances have an indirect stabilizing effect on exchange rate volatility in times when other forms of capital flows are fluctuating drastically by providing a steady source of foreign currency into the dollarized economy. Because economies with high levels of partial dollarization are more sensitive to changes in foreign currency inflows, when diminishing capital flows result in less foreign currency entering the economy, remittances are able to neutralize part of the effect on exchange rate volatility. With sizable remittance inflows, the supplemental source of foreign currency mitigates the size of short term exchange rate movements. In order to test for the relationship between remittances, exchange rate volatility and inflation volatility, I employ a panel analysis using monthly and quarterly data for fifteen developing economies. Using an interaction term between remittances and levels of development, I test whether the marginal effects of remittances are altered by meeting minimum levels of development. I use a similar interaction term between remittances and capital flows to test whether remittances mitigate the effect of net capital outflows on exchange rate volatility. 2

3 2 Literature Review According to the World Bank, remittances are now the second largest resource inflow for developing countries behind foreign direct investment. In 2012, remittances reached $ 350 billion, whereas foreign direct investment flows were $ 616 billion, and foreign portfolio flows were $ 178 billion in low and middle income economies. Remittance inflows have also been more stable than both foreign direct investment and portfolio investment inflows. Remittance flows remain steady for two reasons. First, when remittances enter an economy, they enter as an income transfer from one household (abroad) to another (home country). The transfer of funds can be used for consumption, savings or investment purposes and the funds usually do not leave the economy (Chami, Barajas, Cosimano, Fullenkamp, Gapen and Montiel, 2008). Second, immigrants tend to send a steady flow of remittances and may even increase the amount sent in times of economic downturns in the home country (WorldBank, 2011). The familial relations underpinning the flow of remittances make them unique in their reliability as a stable transfer of funds. As mentioned previously, exchange rate volatility can have severe adverse effects on the domestic economy, trade and investment. The impact on trade and investment activity is rooted in the behavior of risk-averse importers and exporters who are faced with greater uncertainty regarding profitability when exchange rates are volatile. These economic agents will react to volatile exchange rates by decreasing their supply of and demand for goods, in turn adversely affecting economic growth (Sauer and Bohara, 2001). In addition, volatile capital flows that contribute to large fluctuations in exchange rates fuel domestic asset bubbles, poor resource allocation, balance sheet risks and may cause banking or financial crises. Central bank policy choices also depend on the level of dollarization in the economy. In economies that are highly dollarized, such as Peru and Uruguay, Acosta-Ormaechea and Coble (2011) find that inflation targeting is more relevant through the exchange rate channel than through the traditional interest rate channel, linking exchange rate and inflation movements closely together. In contrast, they argue that the interest rate channel is more effective in countries that have less dollarization, such as Chile and New Zealand. The transmission of inflation targeting policies through the exchange rate channel may indicate presence of stronger exchange rate pass-through in dollarized economies. Carranza, Galdon-Sanchez and Gomez-Biscarri (2009) affirm that the negative balance sheet effects derived from depreciation of a currency are more intense in economies with higher levels of dollarization, and that the extent of the pass-through between inflation and exchange rates is significantly affected by the degree of dollarization. Reinhart, Rogoff and Savastano (2003) argue that a higher degree of dollarization in an economy corresponds to a stronger exchange rate pass-through effect, linking volatility in exchange rates to volatility in inflation Carranza et al. (2009) link exchange rates and inflation movements in dollarized economies. They argue that a real depreciation triggers an increase in relative prices of imported goods, leading to greater inflation contingent on the share of imports in consumption spending. They also argue that unique to dollarized economies, deterioration in the balance sheet of firms with high currency mismatch will affect the investment decisions of the firm, and may negatively impact aggregate demand, leading to contractionary effects on output. Because of strong balance sheet effects that cause contractions in times of depreciation, the transfer of exchange rate movements to fluctuations in inflation rates is more profound in partially dollarized economies. Partial dollarization presents numerous risks for the banking system in the form of currency mismatch risk, maturity mismatch risk, and other balance sheet effects. Sudden exchange rate movements can cause bank failures in such cases (Ozsoz, Rengifo and Salvatore, 2010). Banks can self-insure against exposure to currency risk in dollarized economies by engaging in international risk sharing and investing in safe and liquid assets abroad or in foreign reserves at the central bank (Shinichi, 2007). Ozsoz et al. (2010) find that central banks will be more likely to intervene into currency markets when there are higher risks of currency mismatch in partially dollarized economies. 3

4 In addition to the degree of dollarization, the level of development in an economy may explain why some countries fare better than others in the face of greater capital inflows and integration into global capital markets. Kose et al. (2009)) find that countries closer to meeting key thresholds in institutional quality, trade openness and financial system development are able to adapt more effectively to varying inflows. A country s access to global capital market is contingent on its ability to repay its debts, and institutional quality is an important variable in that discussion. Rennhack and Nozaki (2006) find that institutional quality is required to bolster confidence in the domestic economy for countries with high levels of dollarization. Since partial dollarization tends to occur as a reaction to large price fluctuations and has been persistently maintained in many emerging markets, stronger institutions improve faith in the policies of the government, and may diminish the level of dollarization in a country. Honig (2009) finds that stronger control of the exchange rate regime has no impact on the degree of dollarization but better governance decreases partial dollarization when partnered with financial development. The level of trade openness is also a critical factor in realizing gains from capital flows. Trade openness mitigates the risks associated with greater financial integration, particularly in times of economic crises. Kose et al. (2009) argue that trade integration is an essential element in deterring crises related to financial openness and minimizing the costs of such crises if they do occur. Available foreign currency earnings and the ability to maintain foreign-currency-denominated debt obligations make openness to trade important in dollarized economies (Arteta, 2002). Trade helps mitigate crises that stem from currency mismatch risk. When there are sudden stops in capital inflows, the cost of adjustment will depend on the level of trade openness. Frankel and Cavallo (2004) find that in dollarized economies, the size of the domestic contraction in response to a depreciation also depends on the level of trade openness. 3 Greater trade integration corresponds to a smaller contraction in response to the depreciation. Financial development and access to financial services minimize risks associated with greater capital inflows from abroad. Aghion, Bacchetta and Banerjee (2004) find that financial systems which are in the intermediate phase of development tend to be the most unstable in the process of capital account liberalization. Under these circumstances, growth as a result of capital inflows is followed by a collapse or crisis during subsequent outflows. Large fluctuations in capital flows are more difficult for the economy to handle when its financial system is only moderately developed. Acosta, Baerg and Mandelman (2009) argue that an economy with a more developed financial system will experience a smaller appreciation derived from remittance inflows. Financial institutions provide three key services that help minimize the risk of crises. First, strong financial systems alleviate market frictions such as default risk and liquidity constraints by bringing together agents with different time preferences. Second, information and transaction costs are lowered when a strong financial system is able to monitor management and enterprises, as well as screen investment projects. Finally, a developed financial system will improve capital accumulation and productivity growth by efficiently allocating resources to the most productive projects (Beck, 2011). With increased financial openness, strong domestic financial institutions ensure that greater capital inflows to emerging markets do not have adverse effects on the domestic economy, such as exacerbating boom-bust cycles. Aghion and Banerjee (2005) find that when the financial system is weak, changes in flows can intensify domestic economic fluctuations and heighten risk of crises. 3 Frankel and Cavallo (2004) argue that a depreciation of the domestic currency causes a contractionary response in a partially dollarized economy due to the heightened fear that the country will default to avoid a recession causing investors to be more likely to pull out their funds. 4

5 3 Data and Methodology For the main analysis, I use monthly data from 2000 to 2012 for nine developing countries. 4 Monthly data captures the short term dynamics between remittances and volatility measures. Since data on capital flows is not available at a monthly frequency, I use a quarterly data set that includes the nine economies from the monthly analysis as well as an additional six countries. 5 interaction between remittances and capital flows. Next, I separate the countries in each data set into two groups: I use the extended quarterly data set to test the high dollarized and low dollarized economies. To be considered a high dollarized economy, the share of foreign currency deposits in the economy is at least 20 percent of total deposits. Deposit dollarization represents the ability of residents to save and consume in foreign currency. This measure of dollarization addresses how income transfers in the form of remittances can permeate through the economy differently in the two groups. Volatility in the real effective exchange rate are measured as the standard deviation of the log difference for a rolling three-month window. 6 For example, the volatility of exchange rates in March 2012 is the standard deviation of the monthly percent change in exchange rates from January 2012 to March A similar approach is used for inflation volatility, which is calculated as the standard deviation of the inflation rate over a three-month rolling window. This calculation for volatility follows the standard calculation in the literature, and using a rolling window smooths out some of the most extreme fluctuations. In general, remittances have been increasing in absolute size across all economies. In most highly dollarized economies, remittances have also increases in relative size to other resource inflows. A graphical representation of capital flows, remittances, and volatility measures for each economy is presented in the appendix. 7 volatility in exchange rates tends to be greater than volatility in inflation across all countries. Table (1) presents statistical details on real exchange rates and inflation rates from 2000 to 2012 in each country. 3.1 Methodology To analyze the relationship between remittances, exchange rate volatility and inflation volatility, I start with a simple fixed effects model and then expand the model in complexity. 8 The My approach to analysis is threefold. The first model regresses remittances and a set of control variables on inflation volatility and exchange rate volatility respectively. Based on Granger Causality testing 9 I find evidence of dual causality between exchange rate and inflation volatility in highly dollarized economies. To accurately analyze each volatility measure, the regression must account for the exchange rate pass-through while avoiding multicollinearity. The second model employs the residual variables, along with remittances and a set of controls, to test the impact of remittances on exchange rate volatility and inflation volatility separately. Using the residual variables eliminates the likelihood of multicollinearity in the regression while accounting for both the exchange rate pass-through effect in high dollarized economies. The third model employs an interaction term between remittances, net capital flows, and development variables to test whether the influence of remittances on exchange rate volatility changes at a given level of these factors. 4 The nine economies with monthly data are Brazil, Colombia, Dominican Republic, Guatemala, Mexico, Nicaragua, Philippines, and Turkey. Monthly data is sourced from the central banks and statistics offices of each country, the International Financial Statistics database from the IMF and Breugel.org, a think tank tracking real effective exchange rates. 5 The additional economies are Armenia, Georgia, Jordan, Moldova, Poland and Ukraine. 6 As a robustness check, the volatility measure using the standard deviation over a 6 month rolling window is also used for analysis for both exchange rate and inflation volatility. The results did not change. Results are available upon request. 7 Figures (3) and (2) illustrate the inflow of remittances compared to other inflows in each country. Other inflows consist of foreign direct investment and portfolio investment. Figures (5) and (4) present exchange rate and inflation dynamics. 8 I use the Hausman-Wu test to determine that fixed effects is optimal 9 See in Table (9) in the Appendix 5

6 Table 1: Statistical Details for Exchange Rates and Inflation Statistical Summary Real Effective Exchange Rates Inflation Mean Standard Deviation Mean Standard Deviation High Dollarized Armenia % 3.5% 3.2% Dominican Rep % 3.7% 1.6% Georgia % 5.0% 2.3% Guatemala % 1.0% 1.2% Jordan % 1.5% 2.0% Moldova % 2.2% 1.9% Nicaragua % 1.4% 1.8% Philippines % 0.9% 0.9% Turkey % 4.0% 1.6% Ukraine % 2.6% 2.6% Low Dollarized Brazil % 1.0% 0.5% Colombia % 1.0% 0.9% Mexico % 0.7% 0.8% Pakistan % 1.6% 1.7% Poland % 0.8% 0.7% The table uses the extended quarterly data set Model 1: Direct Effects of Remittances on Volatility Measures and Reserve Accumulation The following fixed effects regressions test the direct impact of remittances on inflation volatility and exchange rate volatility respectively. The residuals from each regression will be used in the subsequent models to control for the effects of each factor while avoiding overstating the influence of each variable. V ol.π i,t = f(remit i,t 2, X i,t 2 ) + e π t (1) V ol.rer i,t = h(remit i,t 2, X i,t 2 ) + e r t (2) X i,t = j(t ot i,t, T rade i,t, M2 i,t ) (3) In Equations (1) to (2), e π t and e r t represent the residuals of each regression. The residuals capture the effects of each factor not explained by remittances and the set of controls in X i,t. In Equation (1) the dependent variable V ol.π i,t represents inflation volatility; in Equation (2) the dependent variable V ol.rer i,t represents exchange rate volatility. In addition, Remit i,t measures the remittances as a share of GDP and Equation (3) represents the control variables. The set of controls include terms of trade, T ot i,t, measured as the ratio of the value of exports to the value of imports; the degree of trade openness in the economy, T rade i,t, which is measured as the total value of exports and imports as a share of GDP; and money growth in the economy, M2 i,t, calculated as the monthly percent change in M2. The independent variables are lagged two periods to capture their influence at the beginning of the volatility calculation. Recall that volatility is calculated as the standard deviation over a three month rolling window, therefore the effects of the independent variables are lagged to capture their values at the start of the three month window. The use of residuals in the regression analysis is a common methodology to avoid overstating the influence of specific variables. For instance, by using the residual of inflation volatility in the regression testing the effects of remittances on exchange rate volatility, the effects of remittances and the control variables on 6

7 inflation volatility are stripped away. This concentrates the effect of inflation volatility on exchange rate volatility, which may be necessary when there is high exchange rate pass-through in an economy. This methodology is common in microstructure and health econometrics literature and has been documented by Hasbrouck and Seppi (2001) and Terza, Basu and Rathouz (2008). In the next model, the residuals e π t will be used in the fixed effects regression to understand the relationship between remittances and exchange rate volatility while controlling for the effects of inflation volatility. Likewise, e r t will be used in the regression that tests how remittances impact inflation volatility while controlling for the effects of exchange rate fluctuations on inflation Model 2: Remittances and Exchange Rate Volatility Building in complexity and accuracy from Model 1 above, Model 2 tests how remittances influence exchange rate volatility and inflation volatility respectively. The residuals from Model 1 are incorporated into the analysis to control for exchange rate pass-through between exchange rate and inflation volatility. Model 2 will also incorporate three additional development conditions in the set of controls, which will enter one at a time to test whether they exert an influence on exchange rate and inflation volatility. The development conditions are institutional quality, financial depth and access to financial tools, along with trade openness which is already part of the controls in Model 1. V ol.rer i,t = B 0 + B rm Remit i,t 2 + B π e π t + B z Z i,t 2 + η i,t (4) V ol.π i,t = B 0 + B rm Remit i,t 2 + B r e r t + B z Z i,t 2 + µ i,t (5) Z i,t = k(t ot i,t, T rade i,t, M2 i,t, DC i,t ) (6) Equation (4) tests whether remittances, Remit i,t 2, influence exchange rate volatility while controlling for inflation volatility, e π t, and the set of controls, Z i,t 2. Once again, the independent variables are lagged two periods to capture their effect on volatility at the beginning of the three month window. Equation (5) tests whether remittances influence inflation volatility, while controlling for possible exchange rate pass-through with the residual, e r t, and a set of controls, Z i,t 2. Testing the impact of remittances on price variability while controlling for the effect of exchange rate volatility will provide an additional insight into how dollarized and non-dollarized economies differ in terms of any pass-through effects. Equation (6) defines the set of controls used in Model 2, and differs from the set of controls in Equation (3) by including DC i,t which represents the development conditions. DC i,t will vary depending on the development condition that is being tested. The first of four development conditions being analyzed is institutional quality, Inst i,t. It is measured as the average of five governance indicators from the World Bank Governance Index. 10 The index is collected annually and does not vary drastically in a short period of time. Therefore, given that the institutional factors remain relatively stable on a month-to-month basis, the annual indicators are used as a monthly proxies. To ease comprehension of the results, I assign values ranging from 0 to 5 for the institutional quality variable, instead of the traditional -2.5 to 2.5 that is assigned by the World Bank. The second development condition, Deposits i,t, calculates financial depth as total bank deposits as a percent of GDP. Financial access, Credit i,t, measures the share of claims on the private sector as a share of all domestic claims excluding government. These two variables capture financial development and access to 10 The five indices used in this research are Government Effectiveness, Voice and Accountability, Political Stability and Absence of Violence, Regulatory Quality and Control of Corruption. Figure (6) in the Appendix illustrates the progress in institutional quality for each indicator across the economies. 7

8 credit in the economy. The fourth development condition is trade openness, T rade i,t, and is described above Model 3: Testing Changes in Marginal Effects using Interaction Terms In addition to analyzing the direct effects of remittances on volatility measures, it is also valuable to understand how the marginal effects of remittances change under various conditions. Model 3 uses an interaction term between remittances and net capital flows, as well as remittances and each development condition. The interaction term will capture changes in the marginal effects of remittances on volatility measures. Stabilizing Effect of Remittances First, the interaction between remittances and net capital flows measures whether remittances counter volatility that is derived from large fluctuations in foreign direct investment and portfolio investment. The dependent variable is exchange rate volatility. 11 Diminishing capital flows result in a depreciation of the local currency. A drastic change in capital flows may result in significant fluctuations in exchange rates. Using an interaction term between remittances and net capital flows will capture whether remittances, as a stable and reliable foreign currency inflow, provide a cushion in times of diminishing capital inflows. Net capital flows are measured as the sum of foreign direct investment and portfolio investment flows as a share of GDP. I use the quarterly data set with fifteen economies to conduct the analysis. V ol.rer i,t = B 0 + g(remit i,t, F lows i,t ) + B w W i,t + η i,t (7) g(remit i,t, F lows i,t ) = B rm Remit i,t + B cf F lows i,t + B i Remit i,t F lows i,t (8) W i,t = j(e π t, T ot i,t, T rade i,t ) (9) In Equation (7), Remit i,t F lows i,t measures how the interaction between remittances and net capital flows affects volatility in exchange rates. The following use of coefficients calculates the stabilizing effect of remittances on exchange rate volatility in times of diminishing inflows: [B cf + B i (x)] where (x) represents the level of remittance inflows into the economy. This equation calculates the change in exchange rate volatility derived from a change in capitals flows for a given level of remittances in the economy. The negative sign in front of the calculation addresses the fact that we are considering the impact in times of outflows. Development Conditions I use three alternative interaction terms to test the implications of meeting minimum levels of development on the marginal effects of remittance inflows into dollarized economies. The first is a linear approach that tests whether remittances increase volatility more or less at given levels of development. The next is a nonlinear quadratic approach that examines whether the influence of remittances becomes more or less significant beyond a certain level of development. The last exogenously sets the threshold level of development above the median to test whether a particular level of development matters in quantitative terms The focus in this part of the analysis is on exchange rate volatility. Using the same methodology for inflation volatility yielded results that were not significant. 12 The methodological approach to testing whether threshold levels of development impact the marginal effects of remittances follows the empirical analysis of Kose et al. (2009), who test whether development conditions influence how financial integration affects economic growth. 8

9 Linear Approach The linear analysis examines whether improvements in the development conditions influence how remittance inflows affect exchange rate volatility. The interaction term uses institutional quality, trade openness, financial depth or financial access to test whether the marginal effects of remittances change under varying levels of development. V ol.rer i,t = B 0 + g(remit i,t 2, DC i,t 2 ) + B w W i,t 2 + µ i,t (10) g(remit i,t, DC i,t ) = B rm Remit i,t + B d DC i,t + B rd Remit i,t DC i,t (11) In Equation (10), Remit i,t DC i,t measures how the interaction between remittances and each development condition affects volatility in exchange rates. DC i,t represents each of the four development conditions described in detail above. The lag of two periods captures the impact of the development condition at the beginning of the volatility measure. The set of controls W i,t follows the previous model and is defined by Equation (9). The calculation to test the change in the marginal effects follows: B rm + B rd (y) where (y) represents the level of each development condition being tested. The results will show whether at higher levels of development, remittances contribute more (or less) to volatility in exchange rates. Quadratic Approach The quadratic interaction term tests the non-linear relationship between development conditions and the marginal effects of remittances on exchange rate volatility. The quadratic non-linear equation will capture whether there exist threshold levels of development beyond which the change in volatility derived from remittances is altered. V ol.rer i,t = B 0 + g(remit i,t 2, DC i,t 2 ) + B w W i,t 2 + µ i,t (12) g(remit i,t, DC i,t ) = B rm Remit i,t + B d DC i,t + B rd Remit i,t DC i,t + B dsq DCi,t2 2 + B rdsq Remit i,t DCi,t 2 (13) In Equation (12), the variables are described in the same manner as in Equation (10) and with the same set of controls. The difference comes from the interaction term g(remit i,t, DC i,t ) which now includes a non-linear, quadratic component. The calculation to test the change in the marginal effects follows: B rm + B rd (y) + B rdsq (y 2 ) where (y) represents the level of each development condition being tested. The results will indicate whether a non-linear relationship exists between remittances, the development condition and exchange rate volatility, and whether there is a threshold beyond which the effect of remittances on volatility changes. Median Threshold Approach I use the median of each development condition as an exogenous cut-off value. The exogenous cut off tests whether the development conditions matters only up to a point, and once the threshold is met, the impact on the marginal effects of remittances diminishes. V ol.rer i,t = B 0 + g(remit i,t 2, DC i,t 2 ) + B w W i,t 2 + µ i,t (14) g(remit i,t, DC i,t ) = B rm Remit i,t + B d DC i,t + B dchigh Remit i,t I(DC i,t > DC median ) (15) In Equation (14), the development conditions and control variables are the same as the previous two analyses. The interaction term now includes a dummy variable, I, which captures whether at a given time period the 9

10 economy is above or below the median level of development. If the economy is above the median, I is one, otherwise if it is below the threshold, I is zero. The calculation to test the change in the marginal effects follows: B rm + B dchigh (y) where (y) represents the level of each development condition being tested. The results will capture whether beyond the median level of development, the development condition alters the marginal effect of remittances on exchange rate volatility Robustness Test To test for robustness, I employ three alternative definitions of key variables. First, I use remittances as a share of total inflows. Next, I measure volatility over a six month rolling window. Finally, I test Model 2 using volatility of nominal effective exchange rates 13. The robustness check results for Models 2 and 3 are available in the appendix and are consistent when using the alternative measure of remittance flows, as well as for the six month volatility measure. 4 Results In this section, I present the results derived from the three models discussed above. The first segment addresses the direct effects of remittances on exchange rate and inflation volatility. Next, I present the indirect stabilizing effect of remittances and their ability to mitigate volatility in times of diminishing capital flows. Finally, I discuss the change in marginal effects of remittances under varying levels of development. 4.1 The Direct Effects of Remittance Inflows in Dollarized Economies In highly dollarized economies, an increase in remittance inflows corresponds to an increase in exchange rate volatility and inflation volatility, as can be seen in Table (2). In low dollarized economies, remittances have no direct or statistically significant impact on volatility. The residuals from these regressions are used as controls in subsequent analyses. Table 2: Direct Effects of Remittances on Volatility Measures Dependent Variables: Vol. Real Ex. Rate Vol. Inflation High Low High Low Remittances (% GDP) 0.244*** *** (0.051) (0.185) (0.015) (0.028) Terms of Trade 0.033*** 0.014*** (0.008) (0.005) (0.002) (0.0007) Trade Openness * 0.083*** ** (0.005) (0.021) (0.001) (0.003) M2 Growth ** ** (0.015) (0.021) (0.004) (0.003) R Adj.R N.obs The direct effect of remittances on exchange rate volatility, and inflation volatility is positive for highly dollarized economies. The results are mixed for low dollarized economies. The residuals from these regressions are saved for each group and used as controls in Model 2. Values in parenthesis are standard errors. ***, **, * represent statistical significance of 1%, 5% and 10% respectively. 13 Results available upon request 10

11 Using the residuals derived from the regression presented in Table (2) increases the complexity and strengthens the statistical results that are presented in Tables (3) and (4). Once again, remittances correspond to an increase in volatility in highly dollarized economies. In Table (3), an increase in remittance inflows in highly dollarized economies by one percentage point increases exchange rate volatility by 9 to 210 basis points. The results are robust when controlling for inflation volatility. Improvements in institutional quality and financial access correspond to lower volatility in real exchange rates by 1.6 and 2.1 basis points respectively. Table (4) shows the effect of remittance inflows on inflation volatility. In highly dollarized economies, an increase of one percentage point in remittance inflows corresponds to an increase in inflation volatility of 2.8 to 3.4 basis points. None of the development condition exert a statistically significant effect on inflation volatility. Better institutional quality corresponds to lower inflation volatility in both groups of economies. In low dollarized economies, greater financial depth also corresponds to lower inflation volatility, but neither remittances nor exchange rate volatility influence inflation volatility. A clear relationship between inflation volatility and exchange rate volatility in highly dollarized economies also emerges from this model. It indicates presence of an exchange rate pass-through effect in dollarized economies, as discussed by Carranza et al. (2009) and Reinhart et al. (2003). This relationship is not found in low dollarized economies. Comparing results in Tables (3) and (4) shows that an increase in the inflation volatility residuals corresponds to an increase in exchange rate volatility by 27 to 56 basis points, whereas the increase in the exchange rate volatility residual corresponds to an increase in inflation volatility of approximately 6 basis points. The coefficient for inflation volatility is larger than the coefficient for exchange rate volatility. This may imply a potentially stronger pass-through effect from prices to exchange rates in highly dollarized economies. Table 3: Fixed Effects Analysis: Real Effective Exchange Rate Volatility Dependent Variable: Volatility in Real Effective Exchange Rates High Dollarized Economies Low Dollarized Economies Vol.RER(-1) 0.601*** 0.626*** 0.577*** 0.568*** 0.554*** 0.568*** 0.537*** 0.535*** 0.526*** 0.536*** (17.35) (18.28) (16.12) (15.74) (15.19) (13.71) (12.67) (12.63) (12.33) (12.63) Remittances (% GDP) 0.087*** 0.094** 0.101** 0.102*** 0.098** (2.349) (2.392) (2.482) (2.658) (2.579) (0.708) (0.994) (0.619) (0.917) (0.949) e π i,t 0.505*** 0.268** 0.543*** 0.556*** 0.509*** ( ) (2.251) (4.385) (4.501) (4.132) (0.123) (0.181) (0.063) (0.076) (0.189) Trade Openness ** 0.047*** 0.033* 0.04** (-0.954) (-1.160) (-0.374) (-1.282) (2.500) (2.612) (1.737) (2.313) Terms of Trade 0.012* 0.014* 0.015** 0.015** (1.941) (2.418) (2.538) (2.637) (1.353) (1.241) (1.299) (1.296) M2 Growth (-0.305) (-0.579) (-0.536) (-0.792) (-0.964) (-0.938) (-0.950) (-0.965) Bank Deposits (% GDP) (0.387) (-0.806) Private Credit (% Total) * 0.017* (-1.863) (1.668) Institutional Quality *** (-3.004) (0.075) R Adj.R N.obs Values in parenthesis are standard errors. ***, **, * represent statistical significance of 1%, 5% and 10% respectively. Trade is trade openness, measured as the sum of exports and imports as a share of GDP. The residuals for inflation are e π i,t. An AR(1) term is included to correct for autocorrelation. 11

12 Table 4: Fixed Effects Analysis: Inflation Volatility Dependent Variable: Volatility in Inflation High Dollarized Economies Low Dollarized Economies Vol.Inf(-1) 0.550*** 0.543*** 0.543*** 0.543*** 0.539*** 0.443** 0.433*** 0.421*** 0.431*** 0.424*** (16.56) (16.27) (16.22) (16.26) (16.03) (9.84) (9.565) (9.22) (9.474) (9.256) Remittances (% GDP) 0.034*** 0.031*** 0.028** 0.031*** 0.030*** (3.094) (2.739) (2.401) (2.698) (2.637) (0.595) (0.648) (-0.053) (0.605) (0.104) e r i,t 0.065*** 0.066*** 0.066*** 0.065*** 0.063*** (6.621) (6.695) (6.691) (6.655) (6.236) (0.854) (0.865) (0.672) (0.777) (0.865) Trade Openness * 0.006** ** (0.055) (0.021) (0.171) (0.018) (1.672) (2.009) (1.277) (2.041) Terms of Trade (0.038) (-0.031) (0.036) 0.047) (0.444) (0.224) (0.404) (-0.019) M2 Growth ** ** ** ** (-2.198) (-2.173) (-2.176) (-2.245) (-0.671) (-0.617) (-0.662) (-0.568) Bank Deposits (% GDP) * (0.553) (-1.691) Private Credit (% Total) (-0.315) (0.731) Institutional Quality (-1.209) (-1.303) R Adj.R N.obs Values in parenthesis are standard errors. ***, **, * represent statistical significance of 1%, 5% and 10% respectively. Trade is trade openness, measured as the sum of exports and imports as a share of GDP. The residuals for exchange rate volatility are e r i,t.anar(1)termisincludedtocorrectforautocorrelation. 4.2 Stabilizing Effect: Countering Changes in Other Flows Using quarterly data, Table (5) shows the potential stabilizing effect of remittances. In highly dollarized economies, net capital flows correspond to an increase in exchange rate volatility when the flows are negative, which indicates a decrease or reversal in flows. The bottom segment of Table (5) calculates the interaction term for three levels of remittance inflows: 5, 10 and 25 percent of GDP. As the relative size of remittance inflows increases, exchange rate volatility derived from capital outflows diminishes. The threshold level where remittances exert a stabilizing effect is approximately 20 percent of GDP. As an example, consider two identical countries where the only difference is the amount of remittance inflows. For the country where remittance inflows are only 5 percent of GDP, exchange rate volatility will increase by approximately 4 basis points in times of capital outflows. In contrast, for the country where remittances inflows are 10 percent, under the same capital outflows volatility will increase by only 3 basis points. Beyond the threshold, remittances counteract the impact of the outflow and diminish exchange rate volatility. This in turn provides a stabilizing effect within the dollarized economy. Times of net capital outflows result in a shortage in available foreign currency within an economy. Remittances mitigate the shortage by providing a stable source of foreign currency. They counteract the depreciation pressure and volatility in exchange rates, thereby playing a stabilizing role in periods of capital outflows. For dollarized economies with sizable remittance inflows, central banks need to consider the direct and indirect effects of these inflows when determining intervention into currency markets. Excluding remittances from policy decisions may result in overshooting policy goals related to stabilization. 12

13 Table 5: Stabilizing Effect of Remittances on Real Exchange Rate Volatility Dependent Variable: Volatility in Real Effective Exchange Rates High Dollarized Economies Low Dollarized Economies Vol.RER(-1) 0.329*** 0.332*** 0.324*** 0.225*** 0.208*** 0.198*** (7.911) (7.968) (7.814) (3.703) (3.401) (3.11) Remittances (% GDP) (0.001) (-0.343) (-0.156) (-0.726) (-0.329) (-0.301) Net Capital Flows (% GDP) ** *** ** (-2.538) (-2.711) (-2.356) (0.053) (0.053) (0.122) Remit*NetFlows 0.296* 0.296* (1.709) (1.694) (1.406) (0.080) (-0.096) (-0.104) Trade Openness ** 0.045** (0.975) (1.045) (2.399) (2.412) Terms of Trade (-1.222) (-1.130) (0.352) (0.425) e π i,t 0.149*** (3.687) (0.652) R Adj.R Nobs Calculated Marginal Effects with Net Outflows: [B netflows + B interaction (Remit)] Low Remit (% GDP) Mid Remit (% GDP) High Remit (% GDP) Breakpoint The negative sign in front of the interaction calculation accounts for the effect in times of net capital outflows. Values in parenthesis are standard errors. ***, **, * represent statistical significance of 1%, 5% and 10% respectively. Net Capital Flows represents the net inflows as a share of GDP, excluding remittances. For remittance inflows, the levels are 5%, 10%, and 25 % of GDP for low, mid and high respectively. The residuals for inflation volatility are e π i,t. An AR(1) term is included to correct for autocorrelation. 13

14 4.3 Development Conditions, Remittances and Volatility Measures Institutional quality, financial depth, and trade openness interact with remittances to alter their impact on volatility. Tables (6) through (8) present the three part analysis for each development condition. In each case, the results for the linear, non-linear quadratic, and exogenously set cut-off interaction terms present the dynamics between the levels of development and remittance inflows in both high and low dollarized economies. Institutional Quality The interaction between institutional quality and remittance inflows is summarized in Table (6). The linear analysis indicates that with improvements in institutional quality, the marginal effects of remittances on exchange rate volatility are stronger. The exogenous cut off is also significant, meaning that institutional quality continues to exert an influence on the marginal effects of remittances beyond the median value. The linear and median cut-off effects hold up only to a threshold level of development. For a given amount of remittance inflows, improvements in institutional quality will correspond to lower volatility in exchange rates when institutions are very poor or marginally good. A graphical representation of how institutional quality influences the change in volatility coming from remittance inflows is captured in Figure (1). Economies in the intermediate stage of institutional development will experience an increase in volatility derived from greater remittance inflows. This result follows the discussion presented in Aghion et al. (2004), who argue that countries in intermediate levels of development may actually be the most unstable. Trade Openness Trade openness, like institutional quality, decreases volatility in exchange rates directly and alters the marginal effects of remittances. In Table (7), the linear, non-linear and median cut-off interaction terms all indicate openness to trade mitigates the volatility derived from remittance inflows. The threshold level of trade openness in highly dollarized economies is 36 percent. Before reaching the threshold, greater trade integration corresponds to lower volatility from remittance inflows. Beyond this level, remittances have virtually no impact on exchange rate volatility. This effect can be seen graphically in Figure (1). In low dollarized economies, trade openness is equally significant in the linear analysis, but not in the median cut-off and non-linear quadratic relationship. The results imply that trade integration is important in the relationship between foreign currency inflows and volatility in low dollarized economies. The strength of the influence on volatility does not vary depending on the level of openness. As mentioned previously, trade openness diminishes the economic costs related to exchange rate fluctuations in dollarized economies. Hau (2000) argues that real exchange rate volatility is structurally linked to openness. Recall that trade integration is an important component in mitigating the downside of global financial integration, as found by Kose et al. (2009). From the results in Table (7), it is clear that the benefits of trade openness on stabilizing volatility in exchange rates is not limited to just dollarized economies. Financial System Development Table (8) uses two measures of financial development, financial depth (Deposits i,t ) and financial access (Credit i,t ). In highly dollarized economies, greater financial depth allows remitted funds denominated in either foreign or local currency to enter into the financial system, which may alter the currency mismatch risk on the bank balance sheet. Low levels of financial development exacerbate the risks related to large fluctuations in foreign currency inflows. 14